ChatGPT ads at six months: what the numbers show, and what they don't

OpenAI announced on Monday that its advertising business has reached a $1bn annualised run rate, roughly 200 days after launch. It also opened self-serve access across 31 European markets, plus India, the Middle East and North Africa. Ads now run in more than 40 countries.

In July I suggested UK brands run a small test, and followed it with the test plan I'd hand a client. That advice hasn't changed. What has changed is that we now have six months of advertiser experience to work from rather than a launch announcement, so here's a plain read of where the platform actually is.

What a $1bn run rate means

A run rate is current monthly revenue multiplied by twelve. It describes today's pace, not money already earned. At $1bn, OpenAI is currently taking around $83m a month from ads.

Digiday estimates that translates to roughly $330m in actual revenue booked across the first eight months of the year. OpenAI's stated target for 2026 is $2.5bn in recognised revenue, which would need an average of over $541m a month for the rest of the year.

eMarketer's Nate Elliott called that all but impossible, describing the announcement as "both incredibly impressive and terribly disappointing."

Two useful takeaways for a marketer. The growth is real and fast, which tells you advertiser demand exists. And the headline figure is a forward projection rather than a result, which is worth remembering whenever any platform quotes one.

What advertisers are reporting

The results so far are mixed, and it's worth seeing both sides rather than picking one.

On the positive side. OpenAI's own case studies include an ecommerce advertiser achieving 3x return on ad spend across a 28-day campaign, and a technology partner reporting that more than 80% of its ChatGPT ad traffic came from customers who had never engaged with the brand before. Separately, OpenAI introduced an automated bid strategy called Maximize Results in the second half of August, and one media buyer at Common Thread Collective published account data showing a 108% CTR increase and a 19% lower cost-per-click on the first day of using it.

On the cautious side. Several agencies have published performance data that fell short. Peter Jaffray at Choice OMG spent $415 across three campaigns and recorded 60 clicks at roughly $7 each, a 0.6% click-through rate, and no conversions. Others have reported reasonable click volume with little downstream engagement.

The honest summary is that there's no reliable pattern yet explaining which conditions produce which outcome. OpenAI says so itself, stating in advertiser materials that the platform "does not yet have performance benchmarks across advertisers, industries, or campaign types."

Two things to understand about how it works

Attribution is still maturing. Multiple agencies have reported that click counts in OpenAI's Ads Manager don't reconcile with their analytics. Nicholas Verity of Cleverly recorded 57 clicks in the dashboard against fewer than 20 sessions in Google Analytics from the same campaign. The likely explanation is straightforward: OpenAI's pixel and Conversions API only launched in May, and the platform still lacks view-through attribution, lift measurement and incrementality testing. Those are things every ad platform builds over years. It just means today you should expect the two sources to disagree and plan for it.

Targeting is contextual, not keyword-based. You supply plain-English descriptions of the conversations you want to appear in, and OpenAI matches those semantically against live chats. When it works, it's closer to real intent than a keyword. It isn't precise yet, though. An SE Ranking study of 50,000 US commercial prompts found ads appeared on around a quarter of them, and roughly one in seven of those had no meaningful connection to the conversation they sat beside. Mismatch rates were low in categories like pets and high in relationships and news.

How I'd structure a test

For businesses selling something considered, so travel, finance, software, education or professional services, this is worth a small budget. Four practical points.

Get measurement live before launch. Pixel or Conversions API running, UTMs on every URL, and someone reviewing the dashboard and GA side by side in week one. Decide in advance which source you'll trust when they differ.

Try Maximize Results, but treat it as a hypothesis. The supporting evidence is one buyer's account data rather than a study. It costs nothing to test.

Check where your ads appear. Given the mismatch rate, spend twenty minutes having conversations with ChatGPT in your category. It's the only brand safety check currently available.

Measure at business level. Over four to six weeks, look at enquiry volume and branded search alongside platform reporting. Blunter than you'd accept from an established channel, and right now more dependable.

Where this leaves things

Six months in, ChatGPT ads are a real channel with real reach and self-serve access across most of the markets a UK business would care about. The demand side has proven itself quickly. The measurement side is roughly where you'd expect a platform of this age to be, which is to say usable but not yet trustworthy on its own.

That combination makes it a good candidate for a research budget and a poor candidate for money moved out of a channel that's already working. 🙌

We're testing AI ad platforms with UK clients at the moment, measurement quirks included. Have a chat with us >>